'Monkey Semantics: Two 'Dialects' of Campbell's Monkey Alarm Calls'

Philippe Schlenker, Emmanuel Chemla, Kate Arnold, Alban Lemasson, Karim Ouattara, Sumir Keenan, Claudia Stephan, Robin Ryder, Klaus Zuberbühler

Published version

Preprint version

Errata

(64)b of the published version: 'its source is terrestrial' -> 'its source is non-terrestrial'

Audiovisual resources on Campbell's monkeys (from Zuberbühler's group)

A 2009 BBC report on Ouattara et al. 2009 (with examples)

A linguistics talk on the present research (given as a keynote lecture ('Henry Sweet Lecture') to the annual meeting of the Linguistics Association of Great Britain in 2013)

Q&A on 'Monkey Semantics'

Note prepared by Philippe Schlenker (last update: December 22, 2014)

What is the origin of this research?

In 2010, Schlenker read articles by Ouattara et al. 2009 on the alarm call system of male Campbell’s monkeys. The authors (who became co-authors of the present study) showed that some of these calls display a root-suffix structure (with the roots krak and hok, and the derived forms krak-oo and hok-oo), and that there are some constraints on the syntax of other calls (booms always came at the beginning of sentences). As a semanticist with no prior experience with primatology, Schlenker thought that the data were so rich and interesting that they might benefit from a linguistic analysis. He got in touch with the authors – in particular Klaus Zuberbühler and Alban Lemasson – and after several exchanges a collaboration started.

Zuberbühler and Lemasson’s teams sent to the Paris linguists (Schlenker and Chemla) transcribed sequences of calls obtained in naturalistic situations or in field experiments. These experiments were particularly informative about the meaning of the calls because the triggering situations were known – for instance, some involved playbacks of predator calls (leopard growls and eagle shrieks) with hidden loudspeakers. The Paris linguists worked on theoretical analyses of the data and tested them with quantitative methods, thanks to the expertise of linguist/psycholinguist Emmanuel Chemla and of statistician Robin Ryder.

What are the data discussed in this article?

Ouattara et al. 2009 had rich data on alarm calls produced by male Campbell's monkeys of the Tai forest (Ivory Coast) in naturalistic situations and in field experiments; these data are taken into account in the present study. But we primarily focus on more recent data on Campbell's monkeys collected by Sumir Keenan in the Tai forest (Ivory Coast) and by Kate Arnold on Tiwai island (Sierra Leone) (for comparison, the paper also discusses data on a distinct monkey species, Diana monkeys, obtained from the two sites by Claudia Stephan and Klaus Zuberbühler). Arnold's and Keenan's data made it possible to compare the alarm calls of Campbell's monkeys belonging living in rather different environments. The key environmental difference is that in the Tai forest Campbell's monkeys have two main predators, leopards and eagles; whereas on Tiwai island they are predated by eagles only (leopards haven't been seen for at least 40 years). The article compares the calling behavior of male Campbell's monkeys from the two sites, obtained and transcribed with comparable methods.

What are the main empirical findings?

There are two main empirical findings.

–First, confirming with linguistic means some hypotheses initially made by Ouattara et al. 2009, this article shows that Campbell's calls have a distinction between roots (especially hok and krak) and suffixes (-oo), and that their combination allows the monkeys to describe both the nature of a threat and its degree of danger. For instance, hok warns of serious aerial threats—usually eagles—whereas hok-oo can be used for a variety of general aerial disturbances; in effect the suffix -oo serves as a kind of attenuator.

–Second, the article suggests that one call, krak, is used differently in the Tai forest and on Tiwai island – there is an apparent 'dialectal' distinction between the two sites (to see a somewhat analogous case in human language, speakers of English may think of the word pants, which is used with rather different meanings in the US and in Great Britain). Specifically, krak usually functions as a leopard alarm call in Tai, but as a general alarm call – to warn of all sorts of disturbances, including eagles – on Tiwai. The main argument to come to this conclusion is that in a given type of situation, involving eagle stimuli (for instance eagle shrieks), the monkeys produce different sequences in Tai and on Tiwai: almost no kraks are produced in reaction eagle stimuli in Tai, presumably because krak over there is usually indicative of leopard rather than of eagle presence; by contrast, lots of kraks are produced as a reaction to eagle stimuli on Tiwai island – presumably because there krak functions as a general alarm call.

What are the main theoretical findings?

The article seeks to explain why krak is used in different ways in the Tai forest and on Tiwai island. It explores two theories, with a preferences for the second.

–One theory directly posits that krak has a meaning of general alarm on Tiwai island and a leopard-related meaning in the Tai forest. But the analysis is less simple than one might think, precisely because this 'proto-language' has a bit of structure. Faced with the words krak, hok, krak-oo and hok-oo, a linguist would like to derive the meanings of krak-oo and hok-oo from the meanings of krak and hok, combined with the meaning of -oo. Now as mentioned above, -oo seems to function as a kind of attenuator: hok is primarily used for eagle-related disturbances, whereas hok-oo is also used for less serious aerial disturbances. How can this result be obtained? Suppose hok means: 'there is an eagle-related disturbance'. One possibility is that hok-oo means something like: 'be in the same attentional state as if hok had been uttered' – so: 'look up!' (although this need not be because there is an eagle – that is the difference between hok-oo and hok). On the assumption that in the Tai forest krak means something like 'there is a leopard-related disturbance', we would expect that krak-oo is used whenever one should look down. But in fact even in the Tai forest krak-oo is very clearly used as a general alarm call – it can warn of ground or aerial dangers alike. The authors suggest that these very broad uses of krak-oo indirectly show that even in the Tai forest krak has a 'general alarm' meaning, which enters in the derivation of krak-oo. So the theory ends up being rather complex: on Tiwai island krak has a meaning of general alarm, while in the Tai forest it is ambiguous: it has a leopard-related meaning when krak occurs without a suffix, but in krak-oo it is the general alarm meaning of krak which is used. Linguists don't like to posit ambiguities without reason, and this problem led the authors to explore an alternative theory.

–The authors’ preferred theory is based on the device of ‘implicatures', borrowed from the pragmatics of human languages. It posits that the meaning of a word can be enriched when it competes with a more informative alternative – for instance, “possible” competes with “certain,” which is more informative, and for this reason “possible” usually comes to mean “possible but not certain” (for instance in: “It’s possible that John is the culprit” – which implies that this is not a certainty). The authors propose that krak always has a meaning of general alarm, but that in Tai it comes to be enriched by competition with hok (meaning: aerial alarm) and krak-oo (meaning: non-serious alarm) – with the result that it is enriched with a ‘not hok’ component (hence: the alarm is a non-aerial one) and a ‘not krak-oo’ component (hence: the alarm is a serious one). This yields a meaning of a ‘serious terrestrial alarm’, closely associated with leopards. This operation of 'pragmatic enrichment' fails to arise on Tiwai simply because if it were applied it would yield a useless meaning, as there are no ground predators in that environment.

(To get this analysis to work, the authors give krak a general alarm meaning, they give hok a meaning of aerial alarm, and they take -oo to indicate a non-serious alarm of the type specified by the root. Thus hok-oo is used for non-serious aerial alarms; by competition with hok-oo, hok comes to be used primarily for serious aerial alarms. By the same reasoning, krak-oo warns of non-serious general alarms. The key is that krak competes both with krak-oo and with hok, hence a sophisticated enrichment of its meaning.)

So does this mean that monkey languages can be as sophisticated as human languages? And could Campbell's calls provide clues about the origins of human language?

The paper doesn't in any way imply that monkey languages can be as sophisticated as human languages.(*) First, the root-suffix structure we find in Campbell's calls is very interesting in its own right, but far less complex than what is found in the morphology and syntax of human languages. Second, the mechanism of meaning enrichment used in the authors' preferred theory could be taken as a fairly natural rule in any system that conveys information by way of calls, some of which are more specific than others: if you can use a more specific call, why would you fail to do so?The article shows that as simple as this rule may be, it has far-reaching consequences for the analysis of the meaning of Campbell's calls.

For these reasons, my own position is that it's far too early to tell what, if anything, Campbell's monkey calls can tell us about the origins of human language. I believe that monkey languages are extremely interesting in their own right, and should be the object of detailed studies using the general methods of formal linguistics; but from this it does not follow that the the two kinds of objects (monkey languages vs. human languages) are similar or connected in evolutionary history.

(*) Terminological note: as stated in article, "we freely apply linguistic terminology to these calls" [including the term 'language'], "with the belief that they can and should be studied as formal languages with a sound system, a lexicon, a morphology, a syntax, a semantics, and a pragmatics. Importantly, we do not take a stand on the relation that these systems bear to human language; to say that they can be studied as formal systems does not imply that they share non-trivial properties with human language, nor that they share an evolutionary origin with it."

The article discusses two main theories, and several others are sketched at the end. Is this all sheer speculation?

The article is decidedly theoretical, in that it seeks to offer detailed formal analyses of all the data available. Quite generally in science, several theories are compatible with the available data, although (i) they may incur different costs (e.g. one theory might be much simpler than a competing one), and (ii) they often make different predictions about further experiments that could be undertaken. The case is no different here, but as theoreticians we are very careful to lay out diverse theoretical options (although one theory has our preference). In addition, we are particularly cautious in our conclusions because we are at a very early stage of 'monkey linguistics', and various aspects of the Campbell's monkey calls (notably their syntax, but also aspects of their categorization) are ill-understood.

Why was this article published in a journal called Linguistics & Philosophy? Is this a philosophy paper?

Since our goal is to apply methods from formal semantics to monkey calls, we published our paper in a formal semantics journal. Linguistics & Philosophy is without doubt one of the preeminent semantics journals, with interests in the foundations of the field (including some aspects of the philosophy of language, among others; due its goals and methods, formal semantics has a long tradition of interaction with formal philosophy and philosophical logic). If one looks at the argumentation developed in the article, it is easy to see why it should be assessed semanticists – it used formal tools of that field, and some are non-trivial.

What are the next steps?

We have several works in preparation.

–a comparison of several possible theories of Putty-nosed monkey 'pyow-hack sequences', which were taken in earlier research to be syntactically combinatorial without being semantically compositional (to put it differently: they were taken to have some syntactic structure, but to have a meaning that was 'idiomatic', i.e. not related to this structure); a manuscript is available on LingBuzz;

–a new analysis of complex sequences found in Titi monkeys (a manuscript is ready but not available online);

–a comparison of several possible theories of Black-and-White Colobus sequences, based on data collected by Anne Schel;

–a survey of all our recent results, which should appear in 2015.

What are the long-term perspectives?

In the long term, we hope that this research will help initiate the development of a “primate linguistics” — based on the application of sophisticated methods from contemporary formal linguistics to systems of animal communication. Quite generally, we believe that these methods should make for a much more precise analysis of systems of animal communication – and the data collected by ethologists are now so rich that these methods can definitely help us understand them better.

Press Releases

NYU Press release See also here

CNRS Press release (in French)

Media Coverage [there are various errors and exaggerations in the online media coverage; our article was intentionally modest in its claims (but hopefully ambitious in its methods)]

Open Science World

Scientific American

Notes:

1. The Scientific American article mentions: "Their findings imply that some monkey dialects can be just as sophisticated as human language.". In fact, our article says no such thing, and we definitely do not subscribe at all to this claim.

2. The article mentions that in the Tai forest "krak-oo indicated minor disturbances on the ground", but in fact in the Tai forest krak-oo is used both for ground and non-ground disturbances.

3. A smaller point. The author writes:

“The important thing is that in this situation, both krak-oo and hok are more informative than krak,” he says. “By logic, if you hear krak you can infer there was a reason krak-oo and hok were not uttered, so you infer the negation.”

What is meant is that by the logic of competition (among words), there was a reason krak-oo and hok were not uttered, so you infer the negation.”

Natural News

focus.it [in Italian]

fanpage.it [in Italian]

Le Monde [in French] (see here for a terminological debate on the term 'léopard' [texts in French])

Journal du CNRS [in French]